The proper design of scheduling processes for time-shared computers has provoked much discussion. One of the factors promoting discussion is the great variety of scheduling procedures which one can feasibly program, and which will operate with adequate efficiency. There is wide latitude for conjecturing possibly improved procedures. Moreover, the performance requirements of time-sharing demand more sophisticated schemes than one can analyze with simple queueing theory, yet it is not feasible to experiment with many alternative procedures in operational systems and produce quantitative evidence of their relative merits.